Is the AI content on your website labelled?
We read the provenance data in your images, look for chatbots with no AI notice, and check whether an AI policy is published. Publicly reachable signals only, no sign-up.
- Free
- No sign-up
- Result in ~20 seconds
What gets checked
Image provenance
C2PA manifests and the IPTC DigitalSourceType field: the two formats the AI Act expects machine-readable marking to live in. Plus generator traces that tools like Stable Diffusion leave in the file.
Visible labelling
Is there a notice next to the image that a visitor can read? We check alt text, captions and credit lines in German and English.
Chatbots under Article 50(1)
Which chat vendor is embedded, and is it an AI agent or an inbox staffed by people? A human-staffed live chat triggers no disclosure duty and is never reported as a finding here.
AI policy
Is an AI policy linked, is there an ai.txt or llms.txt? Not required by law, but the usual way to evidence that editorial review takes place.
- It does not detect AI images. No server-side check can. We read only what the generating software wrote into the file.
- Missing marking is not evidence that an image came from an AI. It only shows the marking is missing.
- Invisible watermarks such as Google SynthID cannot be read from outside. That needs the vendor's own verification tool.
- A chatbot's disclosure usually appears only once the widget is opened. We see only the served HTML.
- We check a sample of up to 60 images across several pages, not the whole site.
The legal picture
- Article 50(1): chatbots
- Anyone interacting with an AI system must be told, unless it is obvious. The duty falls on the provider of the system.
- Article 50(2): machine-readable marking
- Outputs of generative systems must be marked machine-readably as artificially generated. This duty falls on the model provider, not on you as a site operator. It becomes your problem when your image pipeline strips the marking back off.
- Article 50(4): your duty as a deployer
- Anyone deploying AI systems under their own authority must disclose deepfakes and label AI-generated text published to inform the public on matters of public interest. Content subject to substantive human editorial review is exempt.
The transparency obligations apply from 2 August 2026. Systems already on the market before that date have until 2 December 2026 to meet the machine-readable marking requirement of paragraph 2. Fines run to EUR 15 million or 3% of total worldwide annual turnover.
Source: Regulation (EU) 2024/1689, Article 50, and the Commission's guidelines on the transparency obligations.
Where does your data go?
The Sovereignty Scan checks the same domain for tools that transfer data to the USA.
Start Sovereignty Scan